JARA-FIT Annual Report 2011. 59. A Bonding Indicator for Plane-Wave. Based Quantum Simulations of Solids. V. L. Deringer1, S. Maintz1, A. L. Tchougréeff1,2, ...
JARA-FIT Annual Report 2011
A Bonding Indicator for Plane-Wave Based Quantum Simulations of Solids V. L. Deringer1, S. Maintz1, A. L. Tchougréeff1,2, R. Dronskowski1 1Institute of Inorganic Chemistry, RWTH Aachen University 2Poncelet Laboratory, Independent University of Moscow, Moscow Center for Continuous Mathematical
Education, Russia Despite its conceptual simplicity and long tradition in the quantum chemistry of solids, a local, orbital-based interpretation of state-ofthe-art density-functional theory (DFT) computations is increasingly difficult, and the reason is simple: today, many DFT simulations rely on plane-wave (PW) basis sets, which describe solids in an efficient but ultimately non-chemical way. To allow for chemical interpretation of such PW-based wave functions, we propose the projected crystal orbital Hamilton population (pCOHP) technique, related to the well-known COOP and COHP approaches.
in bonding or antibonding interaction, one takes the overlap between the two AOs as a measure: a bonding interaction (in the hydrogen molecule, 1+2) will give a positive overlap and thus a positive “crystal orbital overlap population” (COOP). For antibonding interaction, the COOP curve takes negative values, identifying the de‐stabilizing interaction at one glance.
Chemistry, like all sciences, is permanently subject to change—which is advantageous in most cases. Some especially useful concepts, however, have persevered through time: one of them surely is the “chemical bond” formed (in the simplest case) between two adjacent atoms. In the solid state, this notion is especially useful because one has to deal with three‐dimensionally extended crystal structures. And while chemists explore more and more complex solid‐state materials (tailor‐made doped semiconductors, data storage alloys, the next generations of steels, to name but a few), simple, yet powerful chemical‐bonding concepts keep them from getting lost in complexity [1]. Let us start with some analogies. In a molecule, atoms combine their atomic orbitals (AOs) to build the molecular orbitals (MOs), which are usually drawn along an “energy ladder”, and the resulting MO scheme is filled with electrons from the bottom to the top. In molecular hydrogen (H2), the simplest example, two 1s orbitals (denoted ) give one bonding (+) and one antibonding but empty MO (–). In solids, crystal orbitals (COs) result in a comparable way, but in this case, a tiny crystal already contains some 1020 atoms and even more AOs and COs. Luckily, the latter are easily condensed into a densities‐of‐states (DOS) plot which is the solid‐state counterpart of an MO diagram. The DOS indicates where to find electronic states but not their character (are they bonding, antibonding, or do they not contribute at all?). One needs another tool to extract this crucial information. A simple recipe was proposed by Hughbanks and Hoffmann already in 1983 [2]: Draw the bonding DOS contributions to the right side of the energy ladder, and those that are antibonding to the left. To determine if a pair of neighboring orbitals engages
FIG. 1: Schematic depiction of the two routes leading to COHPlike analysis of planewave (PW) computation results. Traditional COHP and novel pCOHP routine arrive at the same result [4]: the filled states in diamond (below F) are all bonding (drawn to the right), while the unoccupied states (above F) are strictly antibonding.
The COOP technique has found many followers, both in chemistry and physics, and it has been built into several codes (often of the semiempirical extended Hückel type). Within DFT, an analogue was proposed in 1993 [3]: the crystal orbital Hamilton population (COHP), which takes the bonding information not from the overlap matrix elements S = but from the Hamiltonian matrix with H = . This time, the expectation value of
59
JARA-FIT Annual Report 2011
the Hamiltonian Ĥ indicates the bond strength. The bottom left part of Fig. 1 shows DOS and COHP plots for diamond, a classical test case for solid‐state theorists. All such orbital‐resolved approaches, however, fail when one uses one of today’s plane‐wave electronic‐ structure codes. The latter expand the crystal’s electronic wavefunction into a large number of plane waves, which has several methodological advantages but practically “abolishes” the chemical picture of atoms and bonds. So far, the usual way out [5] has been to re‐compute the crystal’s wave function using a completely different approach, namely, tight‐binding linear muffin‐tin orbital theory within the atomic‐spheres approximation (TB‐LMTO‐ASA), sketched on the left of Figure 1. To take a different route and extract COHP‐like information directly from a plane‐wave computation, we project the plane‐wave function onto an auxiliary, minimal atomic basis (for the diamond example, one s and three p orbitals at each atom suffice; see references in [4] for details). Once the overlap between the plane waves and an auxiliary orbital is known, we can reconstruct a “projected Hamilton matrix” with the elements H = , and then define the “projected COHP” (pCOHP) in the auxiliary basis, using a simple formulation. Applied to the diamond example, we obtain a qualitatively similar picture directly from the plane‐wave computation, and the resulting pCOHP plot is shown on the right of Fig. 1. Other test cases have been described, including covalently bound, ionic and metallic systems [4]. A more current example is shown in Figure 2. Barium pernitride, BaN2, contains dinitrogen entities in its crystal structure, and chemical intuition suggests a ionic compound to be written as BaN2 = Ba2+ + (N2)2–, so that the pernitride unit will be isoelectronic to molecular oxygen, O=O, and contain a double bond. The bonding in the isolated O=O and (N=N)2– molecules is characterized by a half‐filled set of * orbitals giving a bond order of two [5]. Does this also hold in the solid state? It does, as is easily visualized through the pCOHP shown in Fig. 2. Note that, while the DOS contains all electronic states within the crystal structure, the pCOHP selectively shows the pairwise interaction of the nitrogen valence orbitals, so that the highest (more Ba‐ dominated) bands above 2 eV are visible in the DOS plot but have almost no influence on the N–N pair interaction in the pCOHP. The charge state of the pernitride anion is not too surprising and could be expected; there is no Ba3+ ion. Much to the contrary, transition‐metal cations can exist in different oxidation states: take platinum, which is encountered both as Pt2+ and Pt4+. That being said, platinum pernitride PtN2 can unambiguously be identified as the “+4” compound PtN2 = Pt4+ + (N2)4– [5], and in this case, the * states at nitrogen are completely filled, as a quick look at the N–N COHP of PtN2 affirms [5]. There are many more examples of energy‐partitioning schemes like
60
the COHP providing direct chemical insight into materials of current research interest. Besides the methodological advantage of being based directly on the original PW wave function, where is the advantage of the newly developed pCOHP approach? It is technical but very important.
FIG. 2: A glimpse into recent work [5], reinterpreted using the pCOHP tool. The unit cell of BaN2 is shown together with DOS and –pCOHP plots that have been computed, this time, in a planewave framework (VASP; see references in [4]). The scheme on the right yields the MOs of an isolated “(N2)2–” ion, and the correspondence to the pCOHP is obvious.
The traditional COHP, computed in the tight‐binding approximation, requires densely packed crystal structures throughout. If at all possible, open or lower‐dimensional structures must then be treated by including “empty spheres” [3]. Plane‐wave basis sets, on the other hand, are often the method of choice for supercell computations, allowing facile access to complex scenarios like adsorption on surfaces or a chemical modeling of nanotubes. The pCOHP method promises to extend the range of possible application to many of these systems beyond densely packed solids, and we assume that a lot of interesting chemistry waits to be explored. Accordingly, a program including the pCOHP routines and more advanced projection techniques will be made public in due course. V.L.D. gratefully acknowledges a scholarship from the German National Academic Foundation, and A.L.T. thanks the Russian Foundation for Basic Research for partial support. [1] R. Hoffmann, Angew. Chem. Int. Ed. Engl. 26, 846 (1987). [2] T. Hughbanks, R. Hoffmann, J. Am. Chem. Soc. 105, 3528 (1983). [3] R. Dronskowski, P. E. Blöchl, J. Phys. Chem. 97, 8617 (1993). [4] V. L. Deringer, A. L. Tchougréeff, R. Dronskowski, J. Phys. Chem. A 115, 5461 (2011). [5] M. Wessel, R. Dronskowski, J. Am. Chem. Soc. 132, 2421 (2010).